SoK: Decentralized AI (DeAI)

Zhipeng Wang,Rui Sun,Elizabeth Lui,Vatsal Shah,Xihan Xiong,Jiahao Sun,Davide Crapis,William Knottenbelt

2024-11-26

Abstract:The centralization of Artificial Intelligence (AI) poses significant challenges, including single points of failure, inherent biases, data privacy concerns, and scalability issues. These problems are especially prevalent in closed-source large language models (LLMs), where user data is collected and used without transparency. To mitigate these issues, blockchain-based decentralized AI (DeAI) has emerged as a promising solution. DeAI combines the strengths of both blockchain and AI technologies to enhance the transparency, security, decentralization, and trustworthiness of AI systems. However, a comprehensive understanding of state-of-the-art DeAI development, particularly for active industry solutions, is still lacking. In this work, we present a Systematization of Knowledge (SoK) for blockchain-based DeAI solutions. We propose a taxonomy to classify existing DeAI protocols based on the model lifecycle. Based on this taxonomy, we provide a structured way to clarify the landscape of DeAI protocols and identify their similarities and differences. We analyze the functionalities of blockchain in DeAI, investigating how blockchain features contribute to enhancing the security, transparency, and trustworthiness of AI processes, while also ensuring fair incentives for AI data and model contributors. In addition, we identify key insights and research gaps in developing DeAI protocols, highlighting several critical avenues for future research.

Machine Learning,Artificial Intelligence,Cryptography and Security

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on a series of challenges faced by centralized artificial intelligence (AI) systems, especially those in large - language models (LLMs). Specifically, these challenges include:

1. **Single - point failure**: The control of centralized AI systems is usually in the hands of a single entity, which makes the entire system more vulnerable to technological disruptions or cyber - attacks.
2. **Inherent bias**: Centralized AI systems may reflect the biases of the controlling entity, limiting diverse perspectives, especially in terms of gender, race, socioeconomic status, and geographical representation.
3. **Data privacy issues**: Especially in closed - source large - language models, users' interaction data may be monitored and integrated into the company - controlled model without explicit user consent.
4. **Scalability issues**: As the amount of data and the complexity of tasks increase, the processing capacity of centralized systems becomes a bottleneck, limiting the scalability of the system.
5. **Resource concentration**: Model training resources are concentrated in the hands of a few entities, limiting innovation and hindering the development of the broader AI field.

To address the above challenges, blockchain - based decentralized AI (DeAI) has emerged as a promising solution. By combining the advantages of blockchain and AI technologies, DeAI aims to improve the transparency, security, decentralization, and trustworthiness of AI systems. However, at present, there is still a lack of comprehensive understanding of the latest developments in DeAI, especially the practical solutions in the industry.

Therefore, the goal of this paper is to provide a systematization of knowledge (SoK) of blockchain - based DeAI solutions, propose a taxonomy to classify existing DeAI protocols, and based on this taxonomy, provide a structured way to clarify the current status of DeAI protocols, identify their similarities and differences. In addition, the paper also analyzes the functions of blockchain in DeAI, explores how blockchain characteristics enhance the security, transparency, and trustworthiness of the AI process, while ensuring fair incentives for AI data and model contributors. Finally, the paper points out the key insights and research gaps in developing DeAI protocols, providing directions for future research.